Search Results (3,343)

Urban flexibility research is expanding across buildings, electric vehicles (EVs), distributed energy resources (DERs), storage, positive energy districts (PEDs), digital twins, and interoperability platforms. These strands are often reviewed separately, although urban distribution operators must manage their combined impacts on the same feeders. This paper presents a PRISMA 2020-aligned systematic review with evidence mapping and narrative synthesis of feeder-aware coordination in smart-city electricity systems. Searches of Scopus, Web of Science, IEEE Xplore, ScienceDirect, and citation chasing identified 312 records; 127 studies were included after screening and eligibility assessment, 101 entered the quantitative mapping sample, and 31 formed the deep-synthesis anchor core. Sparse contingency tables were analyzed with Monte-Carlo permutation chi-square tests and bootstrap confidence intervals for Cramér’s V, while ordinal variables were summarized with medians and interquartile ranges. Explicit feeder grounding was concentrated in grid-oriented and EV-oriented studies, whereas many AI/digital-twin and interoperability studies were less often validated against distribution-network operation. Economic and peak-flexibility indicators were reported far more often than interoperability, cybersecurity, or validation-maturity indicators in the anchor core. The synthesis also showed that deployment-oriented work depends on clearer treatment of standards, co-simulation workflows, regulatory instruments, and stakeholder roles. The evidence base is heterogeneous, English-only, and single-coded, so the quantitative results are descriptive rather than population-level. The review contributes a transparent three-layer corpus design (127 included/101 mapped/31 anchor), a domain-specific specialization of SGAM/IEEE 2030 for urban feeder orchestration, an operational digital-twin definition and validation ladder, a retrofittable benchmarking framework, and a practical roadmap for DSOs, municipalities, aggregators, EV operators, building managers, and ICT providers. Full article

Online hate speech and abusive language pose a growing challenge for content moderation, especially in multilingual settings and for low-resource languages such as Lithuanian. This paper investigates to what extent modern multilingual sentence embedding models can support accurate hate speech detection in Lithuanian, Russian, and English, and how their performance depends on downstream modeling choices and feature dimensionality. We introduce LtHate, a new Lithuanian hate speech corpus derived from news portals and social networks, and benchmark six modern multilingual encoders (gemma, qwen, bge, snow, jina, and e5) on LtHate, RuToxic, and EnSuperset using a unified Python pipeline. For each embedding type, we train both a one-class histogram-based anomaly detector (HBOS) and a two-class gradient-boosted tree ensemble (CatBoost), with and without Principal Component Analysis (PCA) compression to 32-dimensional feature vectors. Across all datasets, two-class supervised models consistently and substantially outperform one-class anomaly detection, with the best configurations achieving up to 78.8% accuracy (Kappa 0.58, AUC ROC 0.87) in Lithuanian (jina), 92.2% accuracy (Kappa 0.77, AUC ROC 0.97) in Russian (e5), and 76.9% accuracy (Kappa 0.48, AUC ROC 0.86) in English (e5). PCA compression deteriorates the discriminative power of CatBoost only slightly, with much more negative impact for the HBOS model. These results demonstrate how modern multilingual sentence embeddings combined with gradient-boosted decision trees provides robust machine learning solutions for multilingual hate speech detection applications. Full article

Phishing messages have evolved from simple fraud templates into socially engineered texts that exploit anxiety, trust, relational obligation, and culturally embedded norms. In Korean phishing messages, attackers frequently combine institutional authority, family or acquaintance framing, requests for cooperation, and urgency cues to induce concrete victim actions such as money transfer, link clicking, phone contact, app installation, or credential submission. However, prior studies have largely emphasized binary phishing detection while offering limited interpretability regarding how such messages mobilize social and cultural persuasion strategies. This study proposes a culturally aware large language model framework for analyzing social engineering tactics in Korean phishing messages. The framework is built on a multidimensional codebook that represents the message text, phishing label, tactic type, relation type, requested action, cultural lever, and evidence span, enabling structured and explainable analysis beyond simple classification. To operationalize this framework, an OpenChat-based model is fine-tuned with QLoRA to generate structured outputs that jointly predict the phishing status and socially relevant attributes, while evidence-span supervision is incorporated to improve grounding and explanation consistency. The evaluation examines not only phishing-detection performance but also attribute-level prediction accuracy, evidence alignment, parsing reliability, and human-rated usefulness and trustworthiness. By integrating the cultural context, relational framing, and evidence-grounded explanation into LLM-based phishing analysis, this study provides an interpretable analytical framework for Korean phishing messages and an evidence-grounded basis for analyst-supportive phishing triage. On the 82-sample authoritative clean hold-out split, Model D produced error-free label predictions and achieved 0.841 exact-match core and 0.886 span-F1. However, because the evaluation used a single 82-sample internal hold-out split and no independent external corpus, these results should be interpreted as feasibility evidence under leakage-controlled conditions rather than as proof of deployment-level robustness or cross-domain generalization. The main contribution of this study is therefore not improved binary detection over strong lexical baselines, but the structured and evidence-grounded representation of Korean phishing persuasion tactics for analyst-supportive triage. Full article

The rapid proliferation of online misinformation poses significant risks to democratic processes and public decision-making. However, existing machine learning and deep learning approaches often rely on large annotated datasets and exhibit limited robustness under severe class imbalance and low-resource conditions, particularly in Spanish-language contexts. To address this, this study proposes an interpretable and robust framework for misinformation detection under such constraints. A unified, linguistically grounded and data-centric pipeline is developed, integrating structured lexical, syntactic, and semantic features with synthetic minority augmentation, class-balanced ensemble learning, autoencoder-based representation learning, and active learning under data scarcity. Importantly, the framework systematically evaluates the interaction between these components within a reproducible experimental setting. Results demonstrate that the proposed approach achieves consistent improvements in macro-averaged F1 and minority-class recall compared to baseline models, while reducing performance variance across folds. Ensemble and augmentation strategies provide the most stable configurations, enhancing the detection of underrepresented classes. Moreover, the use of interpretable linguistic features allows predictions to be associated with discourse-level patterns, improving transparency. Consequently, the framework offers a reproducible, computationally efficient, and interpretable solution for misinformation detection in low-resource environments, supporting practical deployment and future multilingual extensions. Importantly, this study provides the first systematic analysis of the interaction between linguistic representations and imbalance mitigation strategies under extreme data scarcity. Full article

Retrieval-Augmented Generation (RAG) systems depend critically on the quality of document preprocessing, yet no prior study has evaluated PDF processing frameworks by their impact on downstream question-answering accuracy. We address this gap through a systematic comparison of four open-source PDF-to-Markdown conversion frameworks, Docling, MinerU, Marker, and DeepSeek OCR, across 21 pipeline configurations, varying the conversion tool, cleaning transformations, splitting strategy, and metadata enrichment. Evaluation was performed using a 50-question benchmark over a corpus of 36 Portuguese administrative documents (1706 pages, ~492K words), with LLM-as-judge scoring over 50 independent runs per configuration. Statistical significance was assessed via Wilcoxon signed-rank tests with Cohen’s d effect sizes. Two baselines bounded the results: naïve PDFLoader (86.2%) and manually curated Markdown (91.3%). Docling with hierarchical splitting and image descriptions achieved the highest automated accuracy (94.1 ± 1.6%), surpassing even manual curation. A per-question-type analysis revealed that table-dependent questions drive the largest accuracy differences, with a 33-percentage-point gap between basic and hierarchical splitting. Metadata enrichment and hierarchy-aware chunking contributed more to accuracy than the conversion framework alone. An exploratory GraphRAG implementation underperformed basic RAG (82% vs. 94.1%). These findings demonstrate that data preparation quality is the dominant factor in RAG system performance. Full article

The incorporation of Artificial Intelligence (AI) into cybersecurity has become widespread, largely propelled by the emergence of Generative AI (GenAI) and Large Language Models (LLMs). While these technologies promise to revolutionize threat detection, they introduce profound challenges regarding explainability, trust, and deployment feasibility in resource-constrained environments. Current research often exhibits a form of technological determinism, prioritizing algorithmic performance over the operational realities of Security Operations Centers (SOCs). This paper presents a hybrid qualitative Systematic Literature Review (SLR) and Mapping Study, adhering to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) 2020 guidelines. Our research questions are narrowly focused, seeking to explore how four key domains intersect: (1) Explainable AI (XAI) methods; (2) cybersecurity operations; (3) human-centered design; and (4) the constraints inherent to edge computing. From an initial corpus of 385 records drawn from Scopus and OpenAlex (spanning a search window from 2014 to 2025, with relevant findings heavily clustered in the 2020–2025 period), included studies were evaluated using a quality assessment protocol adapted from Kitchenham’s guidelines, scoring each study on a 0–24 scale across four dimensions (Venue Quality, Methodological Rigor, Dataset Realism, and Depth of XAI/Human Validation). The results reveal a significant “validation gap”: while 63% of studies claim human-centric relevance, only ~22% incorporate empirical validation with human operators. Furthermore, we identify a critical trade-off between the reasoning power of cloud-based LLMs and the privacy requirements of Edge security. We conclude by proposing a research agenda for “Cognitive SOCs”, emphasizing the need for Small Language Models (SLMs), standardized human-centric metrics, and robust hallucination detection mechanisms. Full article

Background/Objectives: This study describes the development and validation of the Danish Sentence Test (DAST), a Danish-language, adaptive speech-in-noise test constructed from a linguistically balanced corpus using a template-based method. This approach enables controlled linguistic variation while maintaining lexical consistency and may serve as a model for developing similar speech materials in other languages. Methods: Sentences spoken by one female talker from the DAST corpus were sorted into 44 balanced lists of 20 sentences using a psychometric optimization procedure. Speech reception thresholds (SRTs) were measured in 20 normal-hearing participants using headphone playback with speech-shaped noise. Results: Across the 44 sentence lists, the mean SRT was −5.3 dB SNR, with list means within ±0.5 dB of the grand average under the tested configuration. The average within-subject standard deviation was 0.7 dB, and the grand-average psychometric slope was 18.5%/dB. A statistically significant within-session training effect of approximately 0.02 dB per measurement. Conclusions: This study provides normative speech reception threshold (SRT) data for the adaptive Danish Sentence Test (DAST) in normal-hearing listeners under a defined headphone-based speech-in-noise paradigm and demonstrates that the resulting sentence lists yield comparable performance across lists. The template-based construction and optimization approach offers a framework for developing linguistically rich sentence-in-noise tests in other languages. Full article

This article examines the Latin construction i nunc et V_imp from the perspective of diachronic analyzability. The expression consists of two imperative forms with identical morphological marking—the first a motion verb (V1), the second a lexical verb—linked by the conjunction et. Rather than encoding a literal directive sequence, the construction conveys a rhetorical exhortative value that systematically guides discourse interpretation in a direction different from that suggested by its surface form. Although attested from the Imperial period onward, the construction is analyzed against the background of serial imperatives with a motion verb in initial position and verbal pseudocoordination, patterns documented not only in Archaic Latin but also in other historical Indo-European languages. On the basis of an exhaustive corpus, the study assesses the contribution of each constituent in order to account for the construction’s global value. The analysis shows that i nunc et V_imp displays an uneven degree of analyzability: while its components remain formally and syntactically transparent, its semantic and pragmatic analyzability is reduced, as the elements do not contribute compositionally to propositional content but function as a pragmatically unitized block. Overall, the article highlights the central role of analyzability in diachronic change, including processes of unitization and constructional de/recategorization. Full article

The safe operation of power battery systems relies on the dynamic symmetric equilibrium of electrochemical distribution and thermal management states, whereas fault occurrence is often accompanied by symmetry breaking. To achieve accurate fault diagnosis and symmetry restoration, this study proposes a symmetrical closed-loop framework (DBP-KG-FBN) that integrates digital battery passport (DBP) text mining, knowledge graph (KG), and fuzzy Bayesian network (FBN). Power battery fault diagnosis is critical to new energy vehicle (NEV) safety; however, conventional methods face two key limitations: (1) they inadequately exploit multi-source heterogeneous textual data in DBPs; and (2) they fail to handle uncertainty in fault propagation. The methodology proceeds as follows. First, a BERT-BiLSTM-CRF model extracts fault-related entities and relations from unstructured DBP text, which are structured into a Neo4j-based knowledge graph. Second, via rule-based topological mapping, the KG topology is transformed into a Bayesian network through structurally symmetric transformation between the semantic and probabilistic layers, with cyclic dependencies resolved by introducing latent variables. Third, network parameters are determined by integrating fuzzy set theory with game theory-based weighting to quantify uncertainty and subjectivity in expert evaluations, thereby achieving symmetric utilization of subjective and objective information. This enables bidirectional symmetric reasoning for forward fault prediction and backward fault traceability. Experimental results demonstrate that while maintaining symmetric stability of the diagnostic knowledge topology, the proposed DBP-KG-FBN method achieves a diagnostic accuracy of 0.92 (Top-3). This symmetrical closed-loop framework significantly outperforms fault tree analysis (FTA) and event tree analysis (ETA) in diagnostic accuracy and reasoning efficiency. It transforms unstructured DBP data into computable knowledge for intelligent battery diagnosis. Future work will expand the corpus via transfer learning and optimize adaptive weighting algorithms for expert evaluations. Full article

This article investigates the discursive construction of populism in the Portuguese digital public sphere, focusing on the communicative strategies of two party media outlets linked to the populist radical right party CHEGA: Folha Nacional and CHEGA TV. Drawing on Entman’s model of framing functions and the literature on populist communication and digital propaganda, the study examines how these outlets articulate simplified, moralized and emotionally charged narratives to mobilize public opinion and legitimize the party’s political agenda. The empirical corpus consists of 4915 video titles and descriptions published between 2024 and 2025 (CHEGA TV, n = 2476; Folha Nacional, n = 2439). Each unit was coded according to five macro-frames characteristic of populist discourse: (1) appeal to the people and antagonism, (2) messianism, (3) moral restitution, (4) anti-system and anti-elite rhetoric, and (5) exclusion of the other. The research combines qualitative frame analysis with quantitative frequency and co-occurrence analysis, enabling the identification of dominant discursive patterns and their temporal evolution. The study contributes by offering a systematic analysis of populist framing in Chega’s party media, an under-explored field, and by proposing a replicable methodological approach to examine the hybridization of propaganda, emotionality and digital political communication in Europe. Full article

Demyelinating disorders such as multiple sclerosis are characterized by oligodendrocyte loss and insufficient remyelination. The cuprizone model provides a well-established experimental system for studying these processes. The apelinergic system, including the apelin receptor (APLNR), has been implicated in neuroprotection and central nervous system homeostasis. However, its role in white matter demyelination and repair remains incompletely understood. This study aimed to characterize the spatial and temporal dynamics of APLNR expression in relation to oligodendrocyte lineage cells in the corpus callosum (CC) during demyelination and remyelination. Demyelination was induced in 8-week-old C57BL/6 mice by 0.2% cuprizone supplementation in their drinking water for 5 weeks, followed by 5 weeks remyelination phase after toxin withdrawal. Histological assessment using Luxol Fast Blue/Cresyl violet staining was performed to evaluate structural changes in the CC. Immunohistochemistry and confocal microscopy were used to analyze APLNR expression, GST-π⁺ cells, and NG2⁺ cells, including their spatial distribution and co-localization. Quantitative analyses and correlation tests were conducted to assess relationships between cellular markers and CC area. Demyelination resulted in significant reduction in CC area and a marked decrease in GST-π⁺ cells, accompanied by a robust increase in NG2⁺ cells, while remyelination led to partial structural and cellular recovery. APLNR expression increased progressively from control to demyelination and further during remyelination, exhibiting pronounced regional heterogeneity with higher levels in lateral CC regions. Confocal analysis demonstrated increasing co-localization of APLNR with NG2⁺ cells, particularly during remyelination. Correlation analyses identified GST-π⁺ cell density as the strongest predictor of CC area, whereas APLNR showed phase-dependent associations, including a positive correlation with GST-π⁺ cells during remyelination and a negative relationship with NG2⁺ cells during demyelination. APLNR expression is dynamically regulated during cuprizone-induced demyelination and remyelination and is closely associated with oligodendrocyte lineage cell responses. Its increased expression and enhanced co-localization with NG2⁺ cells during remyelination suggest a potential role in endogenous repair processes. However, as the findings are based on descriptive analyses, further functional studies are required to determine the mechanistic contribution of APLNR signaling and its potential as a therapeutic target in demyelinating diseases. Full article

Digital-avatar systems still provide limited control over emotionally expressive behavior in human–computer interaction, especially in Large Language Model (LLM)-based chatbots and virtual assistants with personalized visual embodiments. To address this problem, we propose Multimodal Avatar Generation (MAVAGEN), a multimodal avatar generation framework for synthesizing upper-body digital avatars with personalized appearance and controllable emotional expression. The user specifies the desired gender and age, as well as provides a short text input from which the target emotional state is inferred. MAVAGEN then retrieves an identity image from the HaGRIDv2-1M corpus and generates an avatar clip with synchronized facial expressions, hand gestures, and expressive speech. The framework uses the following six feature streams: textual features, emotion-distribution features, landmark-based pose features, depth-geometry features, RGB-appearance features, and acoustic features. In a quantitative evaluation against recent human animation methods, MAVAGEN achieves the best overall avatar quality, with FID 48.20, FVD 592.00, SSIM 0.741, Sync-C 7.40, HKC 0.929, HKV 25.30, CSIM 0.563, and EmoAcc 0.88. Ablation results show that emotion and acoustic features contribute most to emotional agreement, while landmark-based pose and depth features improve geometric and motion stability. These results support the practical use of MAVAGEN in personalized LLM-based assistants and other emotion-sensitive interactive systems. Full article

Introduction: The growing integration of artificial intelligence (AI) into recruitment and selection is reshaping how organizations identify, evaluate, and choose talent. Although prior research emphasizes improvements in efficiency and automated decision-making, concerns related to fairness, transparency, trust, and applicant experience remain insufficiently resolved. Despite increasing scholarly attention, the field continues to evolve in a fragmented manner. This scoping review addresses this gap by systematically mapping and synthesizing the literature on the advantages and challenges of AI-based recruitment and selection, considering both organizational outcomes and human implications. Materials and Methods: A scoping review was conducted following established methodological frameworks. A structured search and screening process across major academic databases resulted in a final corpus of 33 peer-reviewed studies. The analysis combined descriptive mapping with a hybrid thematic synthesis organized around five dimensions: efficiency and decision support, bias and fairness, transparency and trust, applicant experience, and governance and ethics. Results: The evidence indicates that AI-based recruitment enhances efficiency, scalability, and consistency in decision processes. At the same time, these benefits are accompanied by challenges related to algorithmic bias, limited interpretability, reduced trust, and concerns about procedural fairness. The findings highlight a persistent interdependence between performance outcomes and legitimacy-related responses. Conclusions: This review proposes a socio-technical framework that explains AI-based recruitment as a system shaped by the interaction between technological design, human judgment, and governance structures. The results underscore the importance of integrating oversight, transparency, and ethical accountability to support responsible and sustainable implementation. Full article

Social media sentiment analysis demands systems that are simultaneously accurate, scalable, and interpretable. Lexicon-based methods offer transparency but ignore context, while pre-trained language models (PLMs) capture contextual semantics yet encode sentiment only implicitly. Existing integration strategies inject lexicon signals at the input, attention, or feature layer—all outside the recurrent gating mechanism that controls how affective evidence accumulates over a sequence. We propose the SAGU-Net, a framework built around the Sentiment-Aware Gated Unit (SAGU), a gated recurrent unit (GRU) variant with a dedicated sentiment gate conditioned on external lexicon signals. A complementary Context-Adaptive Sentiment Scoring (CASS) module transforms static polarity scalars into context-dependent vectors via learned projections over PLM representations, bridging the gap between discrete lexicon scores and continuous embeddings. The sentiment gate activations provide token-level explainability without post hoc attribution. On a 12,700-sample Chinese social media corpus of intellectual property co-branding reviews (Fleiss’ $\kappa =0.82$) and two public benchmarks, the SAGU-Net achieves 93.62% accuracy and 93.21% Macro-F1, outperforming nine baselines and matching or exceeding LoRA-fine-tuned large language models (GPT-5, Claude Sonnet 4.6, DeepSeek V3.2, Qwen3.5) while requiring three to four orders of magnitude fewer parameters. Ablation confirms the sentiment gate as the single most impactful component. Full article

In recent years, cyberspace governance has become a critical component of national security strategies worldwide. Although social network platforms provide users with convenient channels for expression and information acquisition, unregulated, harmful euphemisms have become increasingly prevalent. These euphemisms disrupt the order of the digital space and trigger secondary harms such as cyberbullying and regional discrimination. Currently, researches on Chinese harmful euphemism detection face three key challenges: the lack of large-scale annotated datasets, the cognitive reasoning deficit in lightweight models, and the latency constraints of Large Language Models (LLMs), which collectively constrain detection performance and real-world generalization. To address these issues, this study first collected a large corpus from social networking platforms and constructed a fine-grained annotated harmful euphemism dataset. Then, a representation learning framework was designed by integrating deep prompt-based chain-of-thought reasoning with multi-head contrastive learning. This framework introduces external knowledge from LLMs to enhance the diversity and precision of semantic representations. Finally, a multi-dimensional semantic perception fusion framework was proposed. It incorporates multiple semantic perception channels and a cross-channel dynamic fusion mechanism, enabling the model to better capture implicit semantics and integrate external contextual knowledge. Experimental results show that our approach significantly outperforms state-of-the-art lightweight models. While large-scale LLMs exhibit superior zero-shot transferability in cross-domain tasks, our proposed model maintains highly competitive performance with substantially lower inference latency and computational overhead. This research provides a novel methodological and technical foundation for detecting harmful euphemisms in social networks. Full article